Once #552 is closed, we have an endpoint to serve prometheus metrics. We should add more metrics.
For these, we'd want to keep it to something that's actionable. Some ideas from claude:
- Transaction success/failure rate (counter, by status) — "why are 40% of my txs failing?"
- Transaction processing latency (histogram) — "is my program slow or is the RPC client slow?"
- RPC request rate by method (counter) — "is my client spamming getAccountInfo?"
- RPC request latency by method (histogram) — "which RPC calls are slow?"
- Remote account fetch latency (histogram) — "is mainnet cloning the bottleneck?"
Once #552 is closed, we have an endpoint to serve prometheus metrics. We should add more metrics.
For these, we'd want to keep it to something that's actionable. Some ideas from claude: